Readability Pretty

What this tool does

Give it the address of an article and it fetches the page, isolates the main body of text, throws away the navigation, sidebars, cookie banners, newsletter boxes and advertising, and renders what remains in a clean reading view. From there you can export the same article as a PDF.

The result is a single self-contained HTML document shown in a preview frame: a cream page, a serif typeface, generous line height and a measured column width. Images from the article are kept, links stay clickable, and the original URL is printed near the top so you can always trace the source.

How the extraction works

1. The server fetches the page at the URL you supplied. Private and reserved network addresses are rejected by the fetch layer, so the tool cannot be pointed at an internal service.
2. The HTML is passed to a Readability parser, the same family of algorithm used by reader modes in browsers. It scores candidate containers and picks the one that looks most like article content.
3. The chosen content is sanitised: scripts, style blocks and inline event handlers are removed, and only a fixed list of tags survives — paragraphs, headings, lists, blockquotes, code, images, figures and basic inline formatting.
4. Lazy-loaded images are repaired. If an image carries a real address in data-src or data-original, that address is promoted to src, so pictures that would otherwise be blank in a reader often appear.
5. Empty paragraphs, empty spans and long runs of line breaks are cleaned up, and the content is wrapped in a styled document with a title, author, site name, excerpt and source link.
6. That document is placed in the preview frame, and the same HTML is what the PDF export uses.

Extraction is cached for one hour per URL. A repeat request for the same address returns the stored result quickly, which also means an article corrected on the publisher's side may still show the older copy until the cache expires.

What survives and what disappears

Because tables are not on the allowed list, a data-heavy article can lose its tables even though the surrounding prose survives. The same applies to embedded widgets: an interactive chart may leave a gap. Treat the reading view as text-and-image focused, not as a faithful archival copy.

The reading view

The generated page uses a fixed editorial style rather than the publisher's design: a warm off-white background, a white sheet capped at 900 pixels, Georgia with a Times New Roman fallback, and body text at 18 pixels with 1.8 line height. Headings are dark and tight, the excerpt is set as an italic pull quote with a left border, and blockquotes get a tinted background. Images scale to the column width and code is rendered in a monospace stack.

Everything is inline in one document, so the preview does not depend on the publisher's CSS, fonts or trackers. It also means the preview is static: there is no script, no lazy loading and no interactive element.

PDF export

Press Download PDF after a successful extraction and the server regenerates the same article, inlines remote images where it can, and produces an A4 portrait document with 32-point margins. Metadata is written into the file: the article title as the document title, the author when one was detected, and a subject line identifying the export.

The file name comes from the optional PDF Filename field. If you leave it empty, the article title is slugged into a name; characters that are not letters, numbers, hyphens or underscores are replaced. Long articles paginate automatically. Because the PDF is built from the cleaned HTML rather than from the browser preview, fonts are standard PDF fonts, so the serif look of the preview is not reproduced exactly — the text is set in the PDF renderer's default face.

Using the page

1. Paste the article URL. It must include the scheme, for example https://example.com/post.
2. Optionally type a file name for the PDF. This only affects the download, not the extraction.
3. Press Extract. The button shows a spinner while the server fetches and parses the page.
4. Read the Article Info panel: title, author and site badges, a short excerpt and the source link. If the excerpt is missing on the page, the tool generates one from the first 220 characters of the body.
5. Check the preview. Headings, images and quotes should all be present; if something important is missing, the parser probably excluded that container.
6. Press Download PDF to open the export in a new tab. The browser downloads the file rather than displaying it in most cases.

Where it works and where it does not

It works best on ordinary article pages: news, blogs, documentation posts and long-form essays with the text present in the initial HTML. It struggles with pages that render their body with JavaScript after load, pages behind a login or a hard paywall, and sites that block automated requests. In those cases you will see a fetch error, an empty-content error or a preview containing only a teaser.

Infinite scroll, multi-page articles and slideshows are captured as a single page, so only the part present at the URL is included. Comment sections are usually discarded along with other non-article containers, even when they contain useful discussion.

Good uses

• Reading later. Strip the clutter and keep a clean copy of an article you want to read without distractions.
• Printing. The PDF export gives a consistent A4 layout without the publisher's ads.
• Research notes. Save a readable version of a source before it changes or disappears.
• Sharing a quote. The excerpt and clean text are easy to copy into notes.
• Accessibility. A single-column, high-contrast, large-type layout can be easier to read than a busy page.

Limits

One URL per request, no batch mode, no archive of past extractions, no annotations, no text-to-speech and no translation. The tool does not bypass paywalls, does not log in for you and does not execute scripts to reveal hidden content. Extraction results are cached for an hour, and the PDF is generated on demand from the cleaned HTML. Tables, forms and embedded media are outside the allowed tag list and will not appear.

Privacy

The page is fetched by the server, so the publisher sees the server's request, not your browser's. The extracted article is stored in the application cache for an hour to speed up repeats. Do not extract pages that require your session, and remember that the PDF you download contains the article text and any images the server could inline.

Common mistakes

• Passing a bare domain. Use a full URL with https://.
• Expecting tables or videos. They are not in the allowed tag list and are dropped.
• Assuming the preview matches the original design. It is a fixed reading theme, not a copy of the site.
• Re-extracting to get an update. The result is cached for an hour; wait or change the URL.
• Using it on a paywalled article. The tool does not authenticate and cannot reveal gated text.
• Forgetting the file name field. Empty means the title is used as the PDF name.

Frequently Asked Questions